JP2005346419A - Method for processing character and character recognition processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To speed up character recognition processing even when image data are not directed to a normal direction. <P>SOLUTION: An area identification result from an area identification part (12a) is inputted to a direction discrimination part (12b), and the direction discrimination part (12b) successively turns characters to four directions, 0°, 90°, 180° and 270°. A character recognition part (12e) performs character recognition processing for the prescribed number of character patterns in a text area in each direction. The direction discrimination part (12b) judges the direction having the highest recognition rate out of the recognition rates of respective directions as a normal direction and outputs the rotational angle of the normal direction to an image rotation part (12c). At the time of judging that the image data are not directed to the normal direction as the direction discrimination result of the direction discrimination part (12b), the image rotation part (12c) rotates the image data and an area information rotation part (12d) rotates area information (15b). <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a character processing method and a character processing apparatus for processing characters in image data.

  As computers become more sophisticated, document digitization devices such as character recognition devices and electronic filing devices have been put into practical use. In the document digitizing apparatus, image data obtained by reading a document image with a scanner or the like is subjected to character recognition processing, and character data is extracted.

  In the character recognition process, the direction of the original image at the time of reading, that is, the direction of the image data is important. For example, when the original image is in the reverse direction, the characters cut out from the image data are also in the reverse direction. It becomes difficult. Therefore, in the character recognition processing of image data, it is necessary to set the image data in the forward direction.

  Therefore, in the image processing apparatuses disclosed in Patent Documents 1 and 2, a direction determination method for automatically determining the correct direction of image data has been proposed. In these methods, the image data is identified, and the characters in the text region are subjected to character recognition processing in four directions of 0, 90, 180, and 270 degrees, so that the direction with the highest character recognition degree is the normal direction of the document. The image data is rotated based on the direction discrimination result, and the area identification and character recognition processing is executed.

JP-A-8-212298

Japanese Patent Laid-Open No. 2002-024759

  However, in the character recognition processing of Patent Documents 1 and 2, when the image data is not in the normal direction, the region is identified again after the image data is rotationally corrected, so that an accurate region identification result can be obtained. Therefore, the processing speed cannot be sufficiently increased.

  On the other hand, by performing character recognition processing on the image data and embedding the character code of the character recognition result in the image data in an invisible state, when searching by the character code, the image data corresponding to the searched character code is displayed. A technique for creating such character code embedded image data (for example, a technique for creating a PDF data format file by embedding a character code of a character recognition result in image data) is considered. When original image data is displayed in this way, there is an extremely high demand for an improvement in processing speed rather than an accurate region identification result.

  In particular, when processing a large amount of documents at once, such as converting a large amount of documents to image data at once with a scanner equipped with an auto document feeder and creating character code embedded image data by character recognition processing of this image data The processing speed greatly affects the work efficiency.

  The present invention was devised to solve such a conventional problem, and an object of the present invention is to prioritize the processing speed over the accuracy of area identification.

  According to the present invention, even if the image data is not in the normal direction, the area recognition process is performed only once, so that the character recognition process can be speeded up.

  The present invention is a character processing device that performs character recognition processing on image data, and includes image input means for inputting image data, region division of the image data based on attributes, and region information of each region A region identification means for generating a character, and a portion of the region divided region including a character is rotated at a plurality of predetermined rotation angles to perform character recognition processing, and based on the result of the character recognition processing And direction correction means for determining the direction of the image data, and based on the determination result of the direction determination means, the image data is rotationally corrected to any one of the rotation angles as necessary. An image correction unit that performs rotation correction of the region information to any one of the rotation angles, if necessary, based on the determination result of the direction determination unit, and the region information Based in, and a character recognition means for character recognition processing region including the characters. As a result, even when the image data is not in the forward direction, the area recognition process is performed once, so that the character recognition process can be speeded up.

  The present invention relates to a character processing method for performing character recognition processing on image data, an image input step for inputting image data, region division of the image data based on attributes, and region information of each region A region identification step for generating a character, and a portion of the region divided region including a character is rotated at a predetermined plurality of rotation angles to perform character recognition processing, and based on the result of the character recognition processing And a direction determination step for determining the direction of the image data, and based on the determination result of the direction determination step, the image data is rotationally corrected to any one of the rotation angles as necessary. Region correction for correcting the rotation of the region information to any one of the rotation angles, if necessary, based on the image correction step to be performed and the determination result of the direction determination unit And step, based on the region information, and a character recognition step of character recognition processing region including the characters. As a result, even when the image data is not in the forward direction, the area recognition process is performed once, so that the character recognition process can be speeded up.

  Next, preferred embodiments of a character recognition processing apparatus and a character recognition processing method according to the present invention will be described with reference to the drawings.

  In FIG. 1, the character recognition processing apparatus includes an image input unit 11 such as a scanner or a file reading apparatus, and inputs a document as image data from the image input unit 11.

  The image input unit 11 is connected to the bus 17, and is connected to the processor 12 that executes character recognition processing via the bus 17.

  A keyboard 13, a disk 14, a memory 15, and an output unit 16 are further connected to the bus 17.

  The memory 15 stores temporary data when the processor 12 creates processing control information, or stores image data 15 a read by the image input unit 11. In addition, a control program 15c for causing a computer to execute processing corresponding to a flowchart described later is stored.

  The processor 12 executes a control program 15c stored in the memory 15 to thereby execute an area identification unit 12a, a direction determination unit 12b, an image rotation unit 12c, an area information rotation unit 12d, a character recognition unit 12e, and a character recognition result output unit. It functions as 12f.

The area identifying unit 12a identifies the area of the image data 15a and extracts area information 15b. The area information 15b is stored in the memory 15.
The keyboard 13 is a user interface for inputting an instruction from the user and may be any device for inputting an instruction operation from the user, such as a keyboard, a touch panel, or a mouse. Using this keyboard 13, the user can instruct whether to give priority to area identification accuracy or processing speed, and when directed to give priority to area identification accuracy, the document orientation is identified as described in the section of the prior art. Thereafter, the area identification process is performed again. On the other hand, when processing speed priority is instructed, the area identification process is omitted by rotationally correcting the area identification process result performed before identifying the document direction, as described below. Note that when the user instructs the process of automatically converting the input document into the character code embedded image, the processing speed priority is automatically selected. The disk 14 functions as an auxiliary storage device.

  The output unit 16 performs processing of outputting various information such as a display, a printer, a network, a recognition result, and the like, or storing it in another storage device or the like. Further, the output unit 16 may enable data conversion corresponding to the format of the output destination.

  The document read by the image input unit 11 is stored in the memory 15 as image data 15b. The image data 15b is binarized by an image processing unit (not shown) or the like provided in the processor 12 and sent to the region identification unit 12a. The image data 15b is converted to text or graphics by a method such as a histogram method or contour tracking. And is divided into areas such as tables.

  The area identification result of the area identification unit 12a is input to the direction determination unit 12b, and the direction determination unit 12b sequentially rotates the characters in four directions of 0 degrees, 90 degrees, 180 degrees, and 270 degrees for any of the text areas. Here, the character recognition unit 12e performs character recognition processing only for a predetermined number of character patterns in the text area in each direction.

  The direction determination unit 12b determines the direction with the highest recognition rate from the recognition rates in each direction as the positive direction of the document, and outputs the rotation angle in the positive direction to the image rotation unit 12c.

  As a result of the direction determination by the direction determination unit 12b, when it is determined that the image data is not upright, the image data is rotated by the image rotation unit 12c, and the region information 15b is rotated by the region information rotation unit 12d.

  FIG. 3 shows an example in which region identification processing is performed on the image data of FIG. 2 by the region identification unit 12a. 2 includes a title A21 at the top, a text area A22 below the title A21, a text area A23 on the left and right below the text area A22, a figure area A24, a table title area A25 below these areas A23 and A24, A table area A26 exists under the table title A25.

  In FIG. 3, the blocks B31 to B36 corresponding to the areas A21 to A26 are extracted by the area identification process, the block B31 is a text area with a start point (X31, Y31), and the block B32 is a text area with a start point ( X32, Y32), block B33 is a text area, starting point (X33, Y33), block B34 is a drawing area, starting point (X34, Y34), block B35 is a text area, starting point (X35, Y35), The block B36 is a table area and has a start point (X36, Y36). The blocks B31 to B36 are rectangles composed of a horizontal side and a vertical side, and the start point is the coordinates of the upper left corner.

  The region identification result of FIG. 3 is input to the direction discriminating unit 12b, and the direction discriminating unit 12b sets the characters in four directions of 0 degrees, 90 degrees, 180 degrees, and 270 degrees for any of the text areas B31, B32, B33, and B35. The character recognition unit 12e performs character recognition processing in each direction to determine the recognition rate. The character recognition process is executed only for a predetermined number of character patterns in the text area, thereby speeding up the process.

  The direction determination unit 12b determines the direction with the highest recognition rate from the recognition rates in each direction as the positive direction of the document, and outputs the rotation angle in the positive direction to the image rotation unit 12c.

  As a result of the direction determination by the direction determination unit 12b, when it is determined that the image data is not upright, the image data is rotated by the image rotation unit 12c, and the region information 15b is rotated by the region information rotation unit 12d.

  Character recognition processing is executed by the character recognition unit 12e on the forward image data 15a and area information 15b obtained by the above processing, and character data resulting from character recognition is output from the character recognition result output unit 12f.

  FIG. 4 is a diagram showing image data inputted by rotating the image data of FIG. 2 by 180 degrees from the normal direction, and FIG. 5 shows a result of correctly identifying the region of the image data of FIG. In FIG. 5, blocks B51, B52, B53, B54, B55, and B56 correspond to B36, B35, B34, B33, B32, and B31 of FIG. The starting points of the blocks B51, B52, B53, B54, B55, B56 are (X51, Y51), (X52, Y52), (X53, Y53), (X54, Y54), (X55, Y55), (X56, Y56), and is the coordinates obtained by rotating the coordinates of the lower right corner in the corresponding blocks B36 to B31 in FIG. 3 by 180 degrees.

  When the area information in FIG. 5 is rotated in the forward direction by the area information rotating unit 12d, the area information in FIG. 6 is obtained. In FIG. 6, only the position information of the area information is rotationally corrected, and the block number and start point coordinates of FIG. 5 are applied as they are.

  Thus, even in a state in which only position information is rotationally corrected, sufficiently high-precision character recognition is possible.

  Further, the area information rotating unit 12d corrects the block number in the area information of FIG. 6 and matches the start point coordinates with the state in the positive direction to generate the area information of FIG.

  Thus, if the block number and the starting point coordinates are corrected in addition to the position information, the character recognition process can be made more efficient.

  In FIG. 7, the blocks B51, B52, B53, B54, B55, and B56 are corrected to blocks B56, B55, B54, B53, B52, and B51, and the start point coordinates are the coordinates (X71, Y71) of the upper left corner in the positive direction ( X72, Y72), (X73, Y73), (X74, Y74), (X75, Y75), (X76, Y76).

  FIG. 8 shows the area information when the 180-degree rotated image in FIG. 4 is not correctly identified due to the image orientation being incorrect.

  In the area information shown in FIG. 8, the blocks B33 and B35 shown in FIG.

  However, since the two areas (blocks B33 and B35) included in the block 83 are both text areas, the character data is not affected. That is, in the direction determination, it is not necessary to strictly separate regions having the same attribute, and a relatively simple process is sufficient. In this respect as well, the processing is speeded up.

  FIG. 9 shows the result of rotational correction of only the position information based on the result of the direction determination of the region information in FIG. If character regions are recognized in the text regions B83 to B85 and the table region B81 in FIG. 9 and character data is generated, all character data can be acquired correctly.

  FIG. 10 is a flowchart showing the above processing of the character recognition processing apparatus, and shows processing executed when processing speed priority is instructed, such as when the user instructs creation of character code embedded image data. An example of the character recognition processing method implemented by the character recognition processing apparatus is based on this flowchart.

  The processing of the character recognition processing device and the character recognition processing method are executed by the following steps.

  Step S101: First, a document is read from the image input unit 11 and transferred to the memory 15 as image data 15a.

  Step S102: Subsequent to step S101, in order to determine the direction of the image data 15a, an area identification process is performed by the area identification unit 12a, and the areas (image attributes) of the image data are defined as text areas, figure areas, table areas, and the like. ) Is divided into different small areas. In the area identification process, preprocessing such as binarization and inclination correction is also performed.

  The area identification unit 12 a stores area information 15 b (information such as a block number, an attribute, a starting point position, and a size) regarding each area obtained by the area identification processing in the memory 15.

  Step S103: Subsequent to step S102, the direction determination unit 12b rotates the character pattern included in the text region of the region identification result in four directions of 0 degrees, 90 degrees, 180 degrees, and 270 degrees, and for each rotation direction. Character recognition is performed only for a predetermined number of character patterns. This speeds up the process.

  The direction with the highest average similarity (recognition rate) obtained from the result of character recognition is determined as the positive direction, and the rotation angle in the positive direction is output to the image rotation unit 12c.

  Step S104: It is determined whether or not the image data is in the positive direction as a result of the direction determination in step S103. If it is the positive direction, the process jumps to step S107, and if it is not the positive direction, the process proceeds to step S105.

  Step S105: Based on the rotation angle in the positive direction obtained from the determination result of the direction determination, the image rotation unit 12c performs rotation correction processing on the image data 15a so that the image data 15a is in the positive direction. The rotation-corrected image data 15 a is stored in the memory 15.

  Step S106: Based on the rotation angle in the positive direction obtained from the determination result of the direction determination, the region information 15b is rotationally corrected so that the region information 15b is in the positive direction. The rotation-corrected area information 15 b is stored in the memory 15. At this time, only the position information of each block whose rotation has been corrected is corrected for rotation, or the block number and start point coordinates are also corrected.

  Step S107: In the image data 15a that has been rotationally corrected in the forward direction or in the forward direction, a text region is extracted based on the area information 15b that has been rotationally corrected in the forward direction or in the forward direction, and each text region is extracted. Character recognition processing.

  Thereby, character data as a character recognition result is obtained.

  Step S108: Character data obtained as a result of character recognition in step S107 is converted into a predetermined format and output.

  As shown in the above embodiments, even when the image data is not in the forward direction, the processing time can be shortened because the area identification process is performed only once.

  The document recognition processing apparatus is not limited to the configuration shown in FIG. 1, and a general-purpose computer can be applied as the document recognition processing apparatus.

  When the document recognition processing device of the present invention is configured by a general-purpose computer, a computer-executable program including a program code for causing the general-purpose computer to execute each step of the program for the user to execute the character recognition processing is Load it into the computer.

A program for the general-purpose computer to execute document recognition processing is read from a ROM built in the general-purpose computer, a storage medium readable by the general-purpose computer, or read from a server or the like through a network.

It is a block diagram which shows the Example of the character recognition processing apparatus which concerns on this invention. (Example) It is a figure which shows the example of the image data which is a process target of the character recognition processing apparatus of FIG. (Example) It is a figure which shows the result of area | region identification of the image data of FIG. (Example) It is a figure which shows the image data input by rotating the image data of FIG. 2 180 degree | times. (Example) It is a figure which shows the result of area | region identification of the image data of FIG. (Example) It is the figure which carried out the rotation correction | amendment of the area | region information of FIG. 5 to the positive direction. (Example) It is the figure which carried out the rotation correction | amendment of the block number in the area | region information of FIG. (Example) FIG. 5 is a diagram illustrating an example in which the image data in FIG. 4 is not correctly identified in a region. (Example) It is the figure which carried out the rotation correction | amendment of the area | region information of FIG. 8 to the positive direction. (Example) It is a flowchart which shows the Example of the process of the character recognition processing apparatus of FIG. 1, and the character recognition processing method concerning this invention. (Example)

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Image input part, such as a scanner and a camera 12 Processor 12a Area identification part 12b Direction determination part 12c Image rotation part 12d Area information rotation part 12e Character recognition part 12f Character recognition result output part 13 Keyboard 14 Disk 15 Memory 15a Image data 15b Area information 15c Control program 16 Output unit for display, printer, network, etc.

Claims (8)

A character processing device that executes character recognition processing on image data,
Image input means for inputting image data;
A region identification unit that divides the image data based on attributes and generates region information of each region;
Among the divided areas, a part of the area including the character is rotated at a predetermined plurality of rotation angles, character recognition processing is performed, and the direction of the image data is determined based on the result of the character recognition processing. Direction discriminating means for judging;
Based on the determination result of the direction determination means, the image correction means for rotationally correcting the image data to any one of the rotation angles, if necessary,
Based on the determination result of the direction determination means, area information correction means for correcting the rotation of the area information to any one of the rotation angles, if necessary,
Character recognition means for performing character recognition processing on an area including characters based on the area information;
A character recognition processing device characterized by comprising: The character recognition processing apparatus according to claim 1, wherein the rotation angles are 0 degree, 90 degrees, 180 degrees, and 270 degrees. The character processing apparatus according to claim 1, wherein the area information whose rotation is corrected by the area information correcting unit is position information. 4. The character processing apparatus according to claim 3, wherein the area identification means assigns a number to each area, and the area information correction means corrects the number when the area information is rotationally corrected. Furthermore, a selection means for selecting whether to prioritize area identification accuracy or processing speed based on a user instruction;
When the priority of the region identification accuracy is selected by the selection unit, the image data that has been rotationally corrected by the image correction unit is divided into regions based on attributes, and region information for each region is generated. Two-region identification means,
The character processing device according to claim 1, wherein the character recognition unit performs character recognition processing on a region including a character based on the region information generated by the second region identification unit. A character processing method for executing character recognition processing on image data,
An image input step for inputting image data;
A region identifying step for dividing the image data based on attributes and generating region information for each region;
Among the divided areas, a part of the area including the character is rotated at a predetermined plurality of rotation angles, character recognition processing is performed, and the direction of the image data is determined based on the result of the character recognition processing. A direction determining step for determining;
Based on the determination result of the direction determination step, if necessary, the image correction step for rotationally correcting the image data to any one of the rotation angles;
An area information correction step for correcting the rotation of the area information to any one of the rotation angles, if necessary, based on the determination result of the direction determination means;
A character recognition step of performing character recognition processing on a region including characters based on the region information;
A character recognition processing method characterized by comprising: A computer-executable program for realizing the character recognition processing method according to claim 6 by a computer. A storage medium for storing a computer-readable program for realizing the character recognition processing method according to claim 6 by a computer.

Images